https://github.com/nixtla logo
m

marah othman

08/01/2023, 8:47 AM
hello , do I need to make some normalize for my data before using model from neural forecast models ?
m

Mark

08/01/2023, 11:04 AM
this will depend on your dataset and model. I like to run an experiment without, or at least very little, transformations first
m

marah othman

08/01/2023, 12:04 PM
any recommendation to some transformations if you deal with this before?
m

Mark

08/01/2023, 12:04 PM
which model are you using
m

marah othman

08/01/2023, 12:05 PM
i am trying different models neural forecast and mlforecast
i mean when i should make transformations how can i get the knowledge from my data ?this data need or doesn't need like this
c

Cristian (Nixtla)

08/01/2023, 3:40 PM
Hi @marah othman. You dont need to normalize the data before send it to the models. The models of both libraries have inbuilt scalers that take care of the normalization. For
neuralforecast
, you only need to specify the
scaler_type
in the initialization of the model (the recommended scalers are
robust
and
minmax1
m

marah othman

08/02/2023, 2:58 PM
that nice thank you